October 18, 2023
October 18, 2023
Present
Geoffrey Fox, Wes Brewer, Gregor von Laszewski, Christine Kirkpatrick, Gregg Barrett,
Apologies
Sutanay Choudhury, Tom Gibbs, Juri Papay
Tentative Agenda
- Any new members
- White Papers
- Using Benchmarking Data to Inform Decisions Related to Machine Learning Resource Efficiency https://docs.google.com/document/d/1gOKA8BnlJnsTAELWFSmL7Fl7kJej_UrNH-FVXbZFxGI/edit?usp=sharing
- Benchmark Carpentry https://docs.google.com/document/d/15YIlAWOBA2_xjXkTnAZmaw003Jh4eqURVZYQHhdGYdQ/edit#heading=h.fa0u4qc1plw5
- AI Readiness of MLCommons Science https://docs.google.com/document/d/1NbL-VdkrY9jzPxveOys2RCK8TdEJ7O5wgnxjAgzK-rE/edit?usp=sharing
- Other Benchmarks, including Foundation models
- AOB
Discussion
- The power paper is essentially finished and submission to https://www.computer.org/digital-library/magazines/cs/cfp-converged-computing is suggested. Christine contacted editors and they encouraged us.
- We discussed computational load of Foundation models and through Wes found 200M science papers in FORGE https://link.springer.com/article/10.1007/s11227-023-05479-7 and https://drive.google.com/file/d/1ynt_WzjhKBkQSpEQDaSZuMw_zHMyNcgR/view?usp=sharing where second link has computational loads
- Note each AMD GPU on Frontier is two GCD’s where each GCD is a computational unit.
- We noted that Singularity MLCube implementation does not execute on some HPC systems (such as Virginia) due to permission issues
- Foundation model overview https://arxiv.org/pdf/2108.07258.pdf
- We discussed the two BOFs at SC23; Leon Song from DeepSpeed4Science (Microsoft) could speak at one
- We discussed moving to the the benchmark carpentry paper